Syst\u00e9m porovn\u00e1v\u00e1 dvojici sn\u00edmk\u016f takto: p\u0159edtr\u00e9novan\u00e1 s\u00ed\u0165 nalezne koresponduj\u00edc\u00ed body v obou sn\u00edmc\u00edch, ty se filtruj\u00ed pomoc\u00ed geometrick\u00e9 konzistence (homografick\u00e1 verifikace) a v\u00fdsledn\u00e9 sk\u00f3re podobnosti rozhodne, zda jde o stejnou osobu. Kl\u00ed\u010dov\u00fdm krokem je maskov\u00e1n\u00ed oblasti prstu, kter\u00e9 odstran\u00ed artefakty sn\u00edmac\u00edho za\u0159\u00edzen\u00ed. Z p\u011bti testovan\u00fdch architektur (SuperGlue, GlueStick, ASpanFormer, LoFTR, SGM-Net) vynikly zejm\u00e9na LoFTR a ASpanFormer \u2014 oba pracuj\u00ed na b\u00e1zi transformer\u016f bez explicitn\u00ed detekce kl\u00ed\u010dov\u00fdch bod\u016f.<\/p>\n\n\n\n V\u00fdsledky<\/strong><\/p>\n\n\n\n Experimenty prob\u011bhly na t\u0159ech ve\u0159ejn\u00fdch datab\u00e1z\u00edch \u2014 SDUMLA-FV (636 prst\u016f), MMCBNU (600 prst\u016f) a FV-USM (492 prst\u016f) \u2014 s toto\u017en\u00fdm nastaven\u00edm nap\u0159\u00ed\u010d v\u0161emi datasety bez jak\u00e9koli optimalizace.<\/p>\n\n\n\n Ve verifika\u010dn\u00edm sc\u00e9n\u00e1\u0159i (1:1) dos\u00e1hl LoFTR s p\u0159edtr\u00e9novan\u00fdm outdoor modelem na datasetu FV-USM p\u0159esnosti 99,83 % a EER 0,20 %. ASpanFormer dos\u00e1hl na stejn\u00e9m datasetu p\u0159esnosti 99,75 % a EER 0,28 %. Tyto v\u00fdsledky jsou srovnateln\u00e9 se specializovan\u00fdmi modely tr\u00e9novan\u00fdmi p\u0159\u00edmo na biometrick\u00e1 data.<\/p>\n\n\n\n V identifika\u010dn\u00edm sc\u00e9n\u00e1\u0159i s otev\u0159enou sadou (1:N), kde mus\u00ed syst\u00e9m z\u00e1rove\u0148 odm\u00edtnout nezaregistrovan\u00e9 osoby, dos\u00e1hl nejlep\u0161\u00edho v\u00fdsledku dataset MMCBNU s chybovost\u00ed 4,73 %. <\/p>\n\n\n\n Pro\u010d je to d\u016fle\u017eit\u00e9<\/strong><\/p>\n\n\n\n Syst\u00e9m nevy\u017eaduje biometrick\u00e1 tr\u00e9ninkov\u00e1 data, funguje nez\u00e1visle na typu sn\u00edmac\u00edho za\u0159\u00edzen\u00ed a nov\u00e9 osoby lze p\u0159idat za provozu bez p\u0159etr\u00e9nov\u00e1n\u00ed modelu \u2014 co\u017e \u0159e\u0161\u00ed kl\u00ed\u010dovou limitaci sou\u010dasn\u00fdch p\u0159\u00edstup\u016f pro re\u00e1ln\u00e9 nasazen\u00ed s vyu\u017eit\u00edm lok\u00e1ln\u00edch p\u0159\u00edznak\u016f v obraze.<\/p>\n\n\n\n \u0160t\u011bp\u00e1n Rydlo, Filip Ors\u00e1g, Tom\u00e1\u0161 Goldmann, Du\u0161an Kol\u00e1\u0159 Finger Vein Identification Using Pretrained Feature Matching Networks<\/em> IEEE Access, vol. 14, 2026 \u2014 Open Access DOI: 10.1109\/ACCESS.2026.3662722 Podpo\u0159eno FIT VUT Brno, grant FIT-S-23-8151.<\/p>\n","protected":false},"excerpt":{"rendered":" \u017d\u00edly prstu jsou pro biometrick\u00e9 syst\u00e9my atraktivn\u00ed z jednoho prost\u00e9ho d\u016fvodu: jsou skryt\u00e9 pod k\u016f\u017e\u00ed, nelze je okop\u00edrovat otiskem ani fotografi\u00ed a u ka\u017ed\u00e9ho \u010dlov\u011bka jsou jedine\u010dn\u00e9. Modern\u00ed syst\u00e9my pro jejich rozpozn\u00e1v\u00e1n\u00ed dosahuj\u00ed vynikaj\u00edc\u00ed p\u0159esnosti \u2014 ale za cenu specializovan\u00e9ho tr\u00e9ninku pro ka\u017edou novou datab\u00e1zi nebo sn\u00edmac\u00ed za\u0159\u00edzen\u00ed. P\u0159id\u00e1te-li nov\u00e9ho […]<\/p>\n","protected":false},"author":3,"featured_media":1326,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"ub_ctt_via":"","footnotes":""},"categories":[1],"tags":[],"class_list":["post-1323","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-strade"],"featured_image_src":"https:\/\/strade.fit.vutbr.cz\/wp-content\/uploads\/2026\/05\/text392-3-2.png","author_info":{"display_name":"igoldmann","author_link":"https:\/\/strade.fit.vutbr.cz\/index.php\/author\/igoldmann\/"},"_links":{"self":[{"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/posts\/1323","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/comments?post=1323"}],"version-history":[{"count":2,"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/posts\/1323\/revisions"}],"predecessor-version":[{"id":1327,"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/posts\/1323\/revisions\/1327"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/media\/1326"}],"wp:attachment":[{"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/media?parent=1323"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/categories?post=1323"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/strade.fit.vutbr.cz\/index.php\/wp-json\/wp\/v2\/tags?post=1323"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}![\"\"](\"https:\/\/strade.fit.vutbr.cz\/wp-content\/uploads\/2026\/05\/diagram2-1024x357.png\")
<\/figure>\n\n\n\n
\n\n\n\n